A comprehensive test framework to determine the spatial performance of camera-based vehicle detection algorithms

This paper proposes a performance test framework for vision based vehicle detection systems implemented on road-going vehicles. An extensive literature review outlines the evolution of test frameworks used by a number of recently published vehicle detection systems. The proposed test framework determines the effectiveness of a detection algorithm as a function of distance between the host and target vehicles. The framework assists the characterisation of an algorithm´s performance over the full range of distances in which a vehicle can be detected in an image. The test framework is designed for use in blind spot detection, forward collision warning and rear end collision warning applications.